Nobel Lecture: Monetary Neutrality

Hume’s 1752 essays state the quantity theory of money – that changes in the money stock are proportional, neutral “units changes” – while, in the same essays, describing a short-run mechanism by which a monetary expansion “quickens the diligence of every individual” before raising prices, and a contraction induces “poverty, and beggary, and sloth” (pp. 247-248). Lucas presses on the inconsistency: if rational people understand that a monetary injection must eventually raise all prices proportionally, “what force stops this from happening right away?” Hume never explains why early recipients of new money find goods “at the same price as formerly” if the ultimate proportional price rise is, in his words, “easy to trace” and “easily foreseen” (p. 248). Lucas attributes this unresolved tension to the limits of “verbal methods, even someone of Hume’s remarkable powers” (p. 248), and frames the rest of the lecture as an account of the twentieth-century mathematical tools that finally let economists apply Hume’s own “principle of reason” consistently to both the long run and the short run.

Plotting thirty-year (1960-1990) average annual inflation against average M2 growth for 110 countries (from McCandless and Weber 1995) shows the points lying “roughly on the 45-degree line,” with a simple correlation of .95 – rising to .99 for a subset of 14 Latin American countries – while the corresponding plot of money growth against output growth over the same window shows “no relation” between the two (pp. 249-250). Lucas calls the inflation-money relationship a level of empirical success matched by few specific economic theories, and pointedly contrasts it with the absence of “evidence from even one economy” linking interest rates and inflation in a comparably clean way (p. 249). A footnote qualifies the output-growth finding: McCandless and Weber do find a weak positive relation restricted to OECD economies, and other studies have found both signs, so the long-run link between money growth and output growth is “more mixed than one would infer from Figure 2” (p. 250, fn. 2).

Friedman and Schwartz’s (1963) Monetary History shows that “every major depression in the United States over the period 1867-1960 was associated with a large contraction in the money supply, and that every large contraction was associated with a depression,” a correlation Lucas says “gain[s] force from the size of the largest contractions” – an event on the scale of the 1929-1933 Depression being “far beyond anything that can be attributed to shocks to tastes and technology,” leaving monetary contraction as the only candidate shock of adequate size (pp. 251-252). By contrast, Sargent’s (1986) study of the sudden monetary and fiscal reforms that ended four post-World War I European hyperinflations finds reductions in money growth that “dwarf” anything in Friedman-Schwartz, yet these were “not associated with output reductions that were large by historical standards, or possibly by any depressions at all,” because the reforms were credible and so the disinflations were well anticipated (p. 252). The contrast between the two episodes is Lucas’s key empirical illustration that it is the anticipated/unanticipated character of a monetary change, not its size, that governs its real effects.

By the 1960s two incompatible styles of macroeconomics coexisted: general-equilibrium-style monetary theories (Patinkin) that had no operational, testable content, and Tinbergen-style macroeconometric models that could be estimated and simulated but whose relation to microeconomic and classical monetary theory “was unclear” (pp. 253-254). Attempts to build a unified theory by assembling separately-derived intertemporal micro relationships (a consumption function, an investment function, a money-demand function) into a single “church supper” model foundered because those individual decision rules depend on expected future prices, and “the actual equilibrium prices and incomes…bore no relation to, and were in general grossly inconsistent with, the price expectations that the theory imputed to individual agents” (p. 255). Lucas identifies Muth’s (1961) rational-expectations principle – requiring that the expectations built into individual decision rules be consistent with the equilibrium prices those decisions jointly generate – as the fix for this inconsistency, and as the development that “forces the modeler toward a market equilibrium point of view” (p. 255).

Friedman (1968) and Phelps (1968) argued, from general-equilibrium reasoning, that there could be no long-run Phillips-curve tradeoff between inflation and output, yet the standard macroeconometric models of the day embedded exactly such a long-run tradeoff, and the econometric tests in use at the time appeared to reject the natural-rate hypothesis (p. 255). Sargent (1971) and Lucas (1972, 1976) showed that those rejections “depended critically on irrational expectations,” so that once rational expectations was assumed, the existing tests “settled nothing” (p. 255) – reinforcing, from an entirely separate direction, the conclusion that macroeconomics needed a general-equilibrium foundation built on rational expectations.

Two general-equilibrium frameworks were available by the 1960s: the dated-commodity Arrow-Debreu-McKenzie framework used in Kydland and Prescott’s (1982) real-business-cycle model, and Samuelson’s (1958) overlapping-generations “consumption-loan” model; Lucas rules out the former because “such a model without money is obviously not suited to the study of Hume’s problem” (p. 256). In Samuelson’s setup, a young generation can produce but not store goods for its own old age, and an old generation can consume but not produce; absent some institution, the young simply enjoy leisure and never accumulate anything for later. Fiat money can solve this only if the young are willing to accept “intrinsically useless” tokens from the old, trusting they can trade them for goods in their own old age – a possibility Lucas stresses “can certainly not be ruled out,” but is not guaranteed either (pp. 256-257).

With a constant money supply, everyone solves the ordinary problem of choosing labor n to maximize utility from consumption financed by unspent cash, yielding an efficient labor supply n* and a price level p = m/n*, exactly quantity-theoretic and fully neutral (p. 257). Once the money supply grows at a constant rate x via a lump-sum transfer of equal size to every young person, however, solving the first-order condition shows that “the equilibrium level of employment…will…be a decreasing function of the rate of money growth” (p. 258) – the inflation-tax result: because the lump-sum transfer’s value relative to earned cash rises with the rate of money growth, the transfer “dilute[s] the return from working,” and “goods production declines as the inflation rate rises, and everyone is made worse off” (pp. 258-259). Lucas is explicit that this is “obviously not the stimulating effect of a monetary expansion that Hume discusses” – it is a tax effect, not a demand effect – and he removes it from later examples by assuming transfers are instead made in proportion to earnings, restoring full neutrality of the anticipated growth rate (p. 259).

Simply adding uncertainty about the transfer x is not enough, because if the current transfer is revealed “directly only to the old, it will be revealed perfectly to the young by the equilibrium price that it induces” in a single centralized market (p. 260). Real effects require trade to occur outside a single Walrasian market: in Lucas’s (1972) construction, suppliers trade in separate markets with differing numbers of competitors, so that a given price increase can signal either an economy-wide monetary transfer (to be ignored, as a units change) or a market-specific real demand shift (to be met with more output); unable to distinguish the two, suppliers “hedge,” so that “on average…labor supply and production are an increasing function f(x) of the monetary transfer,” with prices completing their proportional adjustment to the transfer only once its true, aggregate size is revealed the following period (pp. 260-261). Lucas frames this and several alternative mechanisms – staggered price-setting (Fischer 1977, Taylor 1979), gradual revelation of shocks (Eden 1994, Williamson 1995) – as different games that all deliver the same qualitative outcome function f(x), noting that “we have no reason to believe that the function f is invariant under changes in monetary policy – it is just a kind of Phillips curve, after all” (p. 261).

Lucas argues that for the purpose of establishing the anticipated/unanticipated distinction it does not matter which game is used, since “any of these models leads to” that same distinction, which he calls “the central lesson of the theoretical work of the 1970s” (pp. 261-262). But he cautions that none of these models derives the outcome function f “from assumptions on technology and preferences alone” – f also depends on stylized, empirically unverifiable assumptions about strategies, timing, and information structure, so “we cannot choose among them on the basis of descriptive realism,” and there is “no reliable way to break [f] down into well understood components” (p. 262).

Tests of Granger-causality (Sargent 1976) and of unanticipated-money residuals (Barro 1977) both find unemployment responding to unanticipated rather than anticipated money, and cross-country tests confirm the model’s prediction that the money-output “multiplier” should shrink as the variance of money shocks rises (Lucas 1973; Alberro 1981; Kormendi and Meguire 1984) (p. 262). However, tests that specifically require monetary shocks to transmit through unexpected price movements are “much less favorable”: estimates in Sargent (1976) and Leiderman (1979) find “only small fractions of output variability…accounted for by unexpected price movements,” leading Lucas to conclude that “though the evidence seems to show that monetary surprises have real effects, they do not seem to be transmitted through price surprises, as in Lucas (1972)” (p. 262) – an explicit acknowledgment that his own original mechanism is not well supported empirically, even though the broader anticipated/unanticipated distinction it helped establish is.

“None of the specific models that captured this distinction in the 1970s can now be viewed as a satisfactory theory of business cycles” (p. 262). He describes much subsequent research as having followed Kydland and Prescott (1982) toward purely real, technology-driven business-cycle models, with more recent work attempting to reintroduce monetary features, a direction in which he expects continued effort. He closes without predicting where the theory will go next, other than that “progress will result from the continued effort to formulate explicit theories that fit the facts” and that the best macroeconomics will keep drawing on developments in basic economic theory (p. 262).